Dynamic split computing for beamforming

Abstract

The present disclosure describes a wireless network that dynamically divides the computation of a beamforming feedback matrix between devices and access points. According to an embodiment, an access point includes an antenna, one or more memories, and one or more processors communicatively coupled to the one or more memories. A combination of the one or more processors determines a first number of layers of a neural network to be implemented by a device and instructs the device to implement the first number of layers of the neural network. The combination of the one or more processors also receives, from the device, intermediate CSI produced by the first number of layers of a neural network implemented by the device, applies, to the intermediate CSI, a second number of layers of the neural network to produce a beamforming feedback matrix, and adjusts the antenna based on the beamforming feedback matrix.

We claim: 
     
         1 . An access point comprising:
 an antenna;   one or more memories; and   one or more processors communicatively coupled to the one or more memories, a combination of the one or more processors configured to:
 determine a first number of layers of a neural network to be implemented by a device based on (i) a size of a channel state information (CSI) matrix for the device and the access point and (ii) computational resources of the device; 
 instruct the device to implement the first number of layers of the neural network; 
 receive, from the device, intermediate CSI produced by the first number of layers of a neural network implemented by the device; 
 apply, to the intermediate CSI, a second number of layers of the neural network to produce a beamforming feedback matrix; and 
 adjust the antenna based on the beamforming feedback matrix. 
   
     
     
         2 . The access point of  claim 1 , wherein the combination of the one or more processors is further configured to:
 change the first number of layers to a third number of layers based on a change in a computational power of the device; and   instruct the device to implement the third number of layers.   
     
     
         3 . The access point of  claim 1 , wherein the first number of layers is determined further based on a transmission rate between the device and the access point. 
     
     
         4 . The access point of  claim 1 , wherein a total number of layers of the neural network is a sum of the first number of layers and the second number of layers. 
     
     
         5 . The access point of  claim 1 , wherein the second number is larger than the first number. 
     
     
         6 . The access point of  claim 1 , wherein the combination of the one or more processors is further configured to transmit a message to the device based on the beamforming feedback matrix. 
     
     
         7 . The access point of  claim 1 , wherein the combination of the one or more processors is further configured to communicate the first number of layers to the device. 
     
     
         8 . A method comprising:
 determining, by an access point, a first number of layers of a neural network to be implemented by a device based on (i) a size of a CSI matrix for the device and the access point and (ii) computational resources of the device;   instructing, by the access point, the device to implement the first number of layers of the neural network;   receiving, from the device, intermediate CSI produced by the first number of layers of a neural network implemented by the device;   applying, by the access point and to the intermediate CSI, a second number of layers of the neural network to produce a beamforming feedback matrix; and   adjusting, by the access point, an antenna based on the beamforming feedback matrix.   
     
     
         9 . The method of  claim 8 , further comprising:
 changing the first number of layers to a third number of layers based on a change in a computational power of the device; and   instructing the device to implement the third number of layers.   
     
     
         10 . The method of  claim 8 , wherein the first number of layers is determined further based on a transmission rate between the device and the access point. 
     
     
         11 . The method of  claim 8 , wherein a total number of layers of the neural network is a sum of the first number of layers and the second number of layers. 
     
     
         12 . The method of  claim 8 , wherein the second number is larger than the first number. 
     
     
         13 . The method of  claim 8 , further comprising transmitting a message to the device based on the beamforming feedback matrix. 
     
     
         14 . The method of  claim 8 , further comprising communicating the first number of layers to the device. 
     
     
         15 . A system comprising:
 a device; and   an access point configured to:
 determine a first number of layers of a neural network to be implemented by the device based on (i) a size of a CSI matrix for the device and the access point and (ii) computational resources of the device; 
 instruct the device to implement the first number of layers of the neural network; 
   wherein the device is configured to:
 apply the first number of layers of the neural network to CSI for a communication channel between the device and the access point to produce intermediate CSI; and 
 communicate the intermediate CSI to the access point; and 
   wherein the access point is further configured to:
 apply, to the intermediate CSI, a second number of layers of the neural network to produce a beamforming feedback matrix; and 
 adjust an antenna of the access point based on the beamforming feedback matrix. 
   
     
     
         16 . The system of  claim 15 , wherein the first number of layers is determined further based on a transmission rate between the device and the access point. 
     
     
         17 . The system of  claim 15 , wherein a total number of layers of the neural network is a sum of the first number of layers and the second number of layers.